No, AI is NOT like the DotCom bubble. Don't believe their B.S.

Overview

The video argues that the current “generative AI” boom is a speculative bubble, and that comparisons to the dot-com era are misleading. The speaker claims that commentators who previously insisted AI wasn’t a bubble are now reframing it as “the next dot-com bubble”—promising future prosperity—but that the underlying mechanics and economic effects are fundamentally different.

Main Claims and Comparisons

• AI looks more like the cryptocurrency bubble than the internet/dot-com bubble.
• Generative AI is “hyped beyond what the tech can deliver.” When the bubble breaks, most people will be shocked by what the technology is actually worth.
• The bubble’s narrative strengthens large incumbent-style players rather than undermining gatekeepers. The speaker argues this increases hype and propaganda compared with the earlier dot-com pattern.
• Investor/executive incentives dominate.
  • If AI delivers productivity gains, everyone could benefit.
  • If it doesn’t, investors and AI providers benefit while workers and most other people lose.
  • The speaker frames the push for AI as self-interested rather than broadly beneficial.

Dot-Com Bubble Framing: Why It Worked (Eventually)

The speaker explains what made the dot-com bubble different:

• The promise (disrupting commerce and enabling new services) was real.
• The main roadblock was a solvable infrastructure bottleneck: the “last mile” problem (connecting homes and enabling efficient logistics).
• After the bubble burst, valuable infrastructure (fiber and interconnection points) remained and was repurposed—eventually enabling the internet people use today.

Cryptocurrency Bubble Framing: Why It Resembles AI

The speaker outlines a similar pattern for crypto:

• The promise was new systems enabling commerce/services without traditional gatekeepers.
• The key practical constraint was network participation requirements (wallets and new infrastructure)—analogous to “last mile” barriers in time and cost.
• However, the winner/loser structure skewed toward companies that only superficially replace intermediaries, especially exchanges and miners, which are positioned to resist disruption.

The speaker then argues that the same winner/loser logic applies to generative AI.

Why the AI “Roadblock” Isn’t Like Dot-Com—and Why Productivity Isn’t Showing Up

The core technical/economic critique:

• Adoption is already high (customers are using AI products), yet productivity gains are not being realized.
• AI companies are losing money, customers complain about AI token costs, and only chip manufacturers appear to be benefiting.
• The speaker argues this is not merely a case of adoption lag due to missing access.
• Instead, the promised gains may not be real—or at least not arriving in any measurable way.

What the Investment Is Actually Building (and Why It May Not “Age Well”)

A major argument concerns how capital expenditures are allocated:

• Most investment is said to fund data centers and AI chips to run larger models.
• If the expected breakthroughs (or “roadblocks”) take time, the speaker argues chips could become quickly obsolete due to ongoing improvements in efficiency.
• Unlike internet infrastructure—where older fiber became more valuable with better protocols—the speaker suggests AI compute capacity tied to new power and chips may not retain value similarly.

Distribution of Benefits: Extractive vs. Expansive

The speaker concludes that generative AI investment is less likely to expand the overall economy and more likely to extract value:

• Generative AI deployments are framed as requiring large centralized data centers.
• This structure enables large companies to charge rents, extracting money rather than creating widespread opportunities for smaller innovators.
• The speaker contrasts this with technologies like the internet, railroads, shipping, airplanes, and telephones, arguing those created broadly distributed economic growth.
• He argues that AI investment, as currently directed, does not “lift all boats” in the same way, but instead consolidates profits toward data-center operators and major AI stakeholders.

Bottom Line

The video concludes that generative AI is a bubble whose long-term economic benefits will be limited, and that harms may also extend to the internet ecosystem—because incentives and infrastructure are pulled toward centralized, rent-seeking models.

The speaker ends by suggesting that AI boosters are effectively selling a “fairy tale,” such as the idea that building massive compute facilities near communities is inherently beneficial.

Presenters / Contributors

• Carl (host/creator; “My name is Carl.”)